Strange 3-3 pattern?

[Downgraded286]

I'm seeing a strange 3-3 combing pattern in these DVD's I've been working with lately. Until now I thought it was either true interlaced, where every frame would have combing... or a 2-3 pattern of 2 interlaced frames and 3 progressive frames. What's up with this wonky 3-3 stuff?

[jagabo]

Pretty much any pulldown pattern is possible. If you want guidance on how to handle it upload a short sample . You can use Dgindex -- mark a short section with motion (~10 seconds) and select File -> Save Project And Demux Video. Upload the resulting m2v file.

[manono]

I'm seeing a strange 3-3 combing pattern in these DVD's I've been working with lately. Until now I thought it was either true interlaced, where every frame would have combing... or a 2-3 pattern of 2 interlaced frames and 3 progressive frames. What's up with this wonky 3-3 stuff?

Nothing 'wonky' about it. As jagabo knows, it's usually indicative of PAL to NTSC, and it can usually be restored to progressive 24.975 or 25fps with the right IVTC. It's easy if you're familiar with AviSynth. But, as he mentioned, to be sure a short untouched sample would be needed.

[Downgraded286]

How strange, when I play the source in MPC-HC none of this is there. Only after running it through MeGui is it present. The combing does appear in DgIndex, but once I save a small clip, then in that clip, too, the combing isn't present. My MPC-HC is set up to show any combing that may exist - deinterlacing is set to Disabled (Progressive) in LAV Video Decoder.

[Downgraded286]

Strange... if I just use QTGMC, there's duplicate frames. If I use TFM and Tdecimate, it seems like there's dropped frames...

[jagabo]

Strange... if I just use QTGMC, there's duplicate frames.

Of course. It's telecined video. That always has duplicate frames after QTGMC.

If I use TFM and Tdecimate, it seems like there's dropped frames...

Because the frame rate isn't 23.976 fps. It's some other frame rate, hence the different pulldown pattern. Any progressive frame rate from 19.98 to 29.97 fps can use pulldown to create the requisite 59.94 fields per second required by DVD. You have to specify the correct settings to TDecimate() to get the original progressive frame rate. If it has blended fields you'll want to use QTGMC() (or Yadif()) and SRestore() instead. But you may still have to specify the correct frame rate.

[manono]

Strange... if I just use QTGMC, there's duplicate frames. If I use TFM and Tdecimate, it seems like there's dropped frames...

Since you ignored the requests (twice) to upload a sample, we can only guess. Assuming it's as you described, try this:

TFM().TDecimate(Mode=0,Cycle=6,CycleR=1)

But without samples no one can be 100% sure. If it's field-blended that's the wrong script to use.

[dphirschler]

I think I saw that once and after analyzing it, I determined it was 25FPS video. I had to write my own script to make it progressive again because it was not a predictable pattern.

Darryl

[jagabo]

if I just use QTGMC, there's duplicate frames.

If you see three copies of every frame after QTGMC, then the original progressive frame rate was 1/3 of the QTGMC video's frame rate. This would typically happen with 8mm film, sped up to 19.98 fps, then telecined.

[Downgraded286]

Strange... if I just use QTGMC, there's duplicate frames. If I use TFM and Tdecimate, it seems like there's dropped frames...

Since you ignored the requests (twice) to upload a sample, we can only guess. Assuming it's as you described, try this:

TFM().TDecimate(Mode=0,Cycle=6,CycleR=1)

But without samples no one can be 100% sure. If it's field-blended that's the wrong script to use.

I apologize for that, but as I said before, for some reason in the original sample I saw no combing... only after transcoding it. I guess I could have made a sample of that, but I digress.

In any case, that code seems to have worked. What exactly does that code do, anyway? It resulted in a 24.975FPS video... is this indicative of a PAL-NTSC crossover?

[manono]

What exactly does that code do, anyway?

It performs an IVTC. Instead if doing it for the more common 3:2 pulldown, with those settings it undoes 3:2:3:2:2 pulldown.

...is this indicative of a PAL-NTSC crossover

Yes.

I guess I could have made a sample of that...

We were asking for a short sample straight from the DVD. The excuses given for not doing as requested make no sense.

[jagabo]

I apologize for that, but as I said before, for some reason in the original sample I saw no combing...

Then upload a sample of that.

only after transcoding it. I guess I could have made a sample of that

Are you looking for real help or not? Without seeing samples nobody can help you. We can only make guesses from the sparse (and probably unreliable) information you have given.

In any case, that code seems to have worked. What exactly does that code do, anyway?

It field matches frames (making progressive frames whenever possible) and then discards one (hopefully a duplicate) of every six frames.

It resulted in a 24.975FPS video... is this indicative of a PAL-NTSC crossover?

Maybe. But without a sample of your original source nobody can tell exactly what's going on.

If you had uploaded some samples this would have been solved a long time ago and you might have learned something.

[U2Joshua]

Okay, same situation, here: 3:3 Pulldown?

Sample: https://www.sendspace.com/file/sjyi8x

Is @manono's script [ TFM().TDecimate(Mode=0,Cycle=6,CycleR=1) ] the best recommendation?

Like OP I was formerly familiar with 3:2, but hadn't yet encountered 3:3.

thanks in advance~

[manono]

Is @manono's script [ TFM().TDecimate(Mode=0,Cycle=6,CycleR=1) ] the best recommendation?

It's not really my script, but yes.

Like OP I was formerly familiar with 3:2, but hadn't yet encountered 3:3.

As I mentioned in my previous post, it's 3:2:3:2:2 pulldown. Bob it and you'll see. Where, with the standard 3:2 pulldown you'll have a pattern of repeated frames (after bobbing) of 3, 2, 3, 2, 3, 2, with 3:2:3:2:2 pulldown you'll get 3, 2, 3, 2, 2. 3:3 pulldown is something else again.

[U2Joshua]

Is @manono's script [ TFM().TDecimate(Mode=0,Cycle=6,CycleR=1) ] the best recommendation?

It's not really my script, but yes.

Like OP I was formerly familiar with 3:2, but hadn't yet encountered 3:3.

As I mentioned in my previous post, it's 3:2:3:2:2 pulldown. Bob it and you'll see. Where, with the standard 3:2 pulldown you'll have a pattern of repeated frames (after bobbing) of 3, 2, 3, 2, 3, 2, with 3:2:3:2:2 pulldown you'll get 3, 2, 3, 2, 2. 3:3 pulldown is something else again.

Hmm... by "Bob (it)" do you simply mean to do a frame-by-frame manual inspection of the clip to discern the interlace pattern? If so, I've done that, and I think what I'm observing is 3 interlaced frames to 3 progressive frames, in repeating sequence.

[manono]

Hmm... by "Bob (it)" do you simply mean to do a frame-by-frame manual inspection of the clip to discern the interlace pattern?

No. To 'bob' means to create full frames from fields, thus doubling the framerate to 59.94fps. You can put on the built in AviSynth filter Bob (and you'll learn why it's called 'bob' as you notice the frames created bouncing up and down), or you can use Yadif(Mode=1) or put on QTGMC which bobs by default. Or a number of other bobbing filters

[U2Joshua]

Hmm... by "Bob (it)" do you simply mean to do a frame-by-frame manual inspection of the clip to discern the interlace pattern?

No. To 'bob' means to create full frames from fields, thus doubling the framerate to 59.94fps. You can put on the built in AviSynth filter Bob (and you'll learn why it's called 'bob' as you notice the frames created bouncing up and down), or you can use Yadif(Mode=1) or put on QTGMC which bobs by default. Or a number of other bobbing filters

I̶ ̶s̶e̶e̶.̶.̶.̶ ̶ ̶s̶o̶,̶ ̶o̶n̶c̶e̶ ̶I̶ ̶d̶o̶ ̶"̶B̶o̶b̶"̶ ̶i̶t̶,̶ ̶h̶o̶w̶ ̶t̶h̶e̶n̶ ̶d̶o̶ ̶I̶ ̶i̶n̶t̶e̶r̶p̶r̶e̶t̶ ̶t̶h̶e̶s̶e̶ ̶p̶u̶l̶l̶ ̶d̶o̶w̶n̶ ̶p̶a̶t̶t̶e̶r̶n̶s̶?̶ ̶ ̶C̶o̶u̶n̶t̶,̶ ̶f̶r̶a̶m̶e̶-̶b̶y̶-̶f̶r̶a̶m̶e̶,̶ ̶a̶n̶d̶ ̶l̶o̶o̶k̶ ̶f̶o̶r̶ ̶d̶u̶p̶l̶i̶c̶a̶t̶e̶ ̶f̶r̶a̶m̶e̶s̶?̶ ̶ ̶S̶o̶r̶r̶y̶.̶.̶.̶ ̶ ̶I̶'̶m̶ ̶s̶u̶r̶e̶ ̶m̶u̶c̶h̶ ̶o̶f̶ ̶t̶h̶i̶s̶ ̶i̶s̶ ̶v̶e̶r̶y̶ ̶b̶a̶s̶i̶c̶.̶ ̶ ̶I̶ ̶a̶p̶p̶r̶e̶c̶i̶a̶t̶e̶ ̶y̶o̶u̶r̶ ̶p̶a̶t̶i̶e̶n̶c̶e̶ ̶a̶n̶d̶ ̶h̶e̶l̶p̶.̶

Edit: Okay... I bobbed the clip using QTGMC and sure enough I now understand what you're saying. I see the 3:2:2:3:2 pattern.

That said, is it just something that folks like you (who have much more experience) know what script(s) to use?

Further, is this practice of Bobbing a good starting point for determining pulldown patterns in the future? I have a lot more videos with seemingly unusual patterns. If you say so, I'll bob them, count the pattern, and then at least have a better starting point in asking for assistance on here from you veterans.

[manono]

When dealing with something out of the ordinary, I almost always begin by bobbing it with Yadif(Mode=1). Yadif is fast and I can tell what's going on much more easily than I can with QTGMC, although I'll often use QTGMC for the actual encoding when bobbing is needed before unblending, or (more often these days), yadifmod(mode=1, edeint=nnedi3(field=-2)) (which needs YadifMod and NNEDI3) followed by SRestore(), because it's faster than QTGMC, if not quite as good.

With PAL to NTSC (as yours is) more often than not the conversion involves field blending and bobbing shows the blended fields quite nicely. Then you bob it followed by SRestore at default settings. Sometimes a field blended source can be film and it was field blended to 29.97fps. Then you'd bob it followed by SRestore(Frate=23.976).

Your 6-frame cycle of the unfiltered source is a giveaway that it's PAL to NTSC. And almost a giveaway that it can be IVTC'd. Bobbing it and counting the 3, 2, 3, 2, 2 pattern without any field blending confirms the theory and tells you it can be IVTC'd with those settings given earlier. Field blended PALtoNTSC generally (not always) has 5 out of 6 frames interlaced.

So, once you decide it's not been field blended you can put on:

TFM()

all by itself. Then count how often you see duplicate frames. Every six frames and it's 3:2:3:2:2, every 5 frames and it's 3:2. And you follow TFM with TDecimate at the correct settings. There are lots of ways to figure out what's going on (and others I haven't mentioned) and you'll learn from experience if you're unfortunate enough to have to deal with screwy sources often.

[jagabo]

Think about what the 3:2:2:3:2 pattern means: 5 different film frames were converted into 12 fields. Since the field rate of the NTSC video was 59.94 the progressive frame rate was 59.94 * 5 / 12 = 24.975 frames per second. The last issue is whether the 25 fps source was slowed to 24.975 fps before being telecined (almost never) or whether the 3:2:2:3:2 pattern was modified occasionally (most likely) to make up for the difference. It's almost impossible to tell the difference and, in practice, it's not worth worrying about.

[U2Joshua]

That does make sense, @jagabo. Thank you, and thank you, @manono.

How about 25fps PAL DVDs that are known to be from 24fps film sources? I can't quite wrap my head around this one, yet... Lots of conflicting information online, it seems. Here, in this example, there are far more interlaced frames (maybe all?) than progressive.

SAMPLE: https://www.sendspace.com/file/5659us

thanks in advance~

[manono]

It was field blended so you unblend it and get back 23.976fps:

QTGMC()###or your bobber of choice
Srestore()

Very common, that kind.

Lots of conflicting information online, it seems.

Don't know about that. There are several ways to take a film source and convert it to PAL. Field blending is one of the worst. But it's easy so it's used some of the time.

[U2Joshua]

It was field blended so you unblend it and get back 23.976fps:

QTGMC()###or your bobber of choice
Srestore()

Very common, that kind.

Lots of conflicting information online, it seems.

Don't know about that. There are several ways to take a film source and convert it to PAL. Field blending is one of the worst. But it's easy so it's used some of the time.

Thanks, @manono... testing this out, and it's working beautifully.